Well drilling and completion operation site environment risk early warning method

Abstract

The invention discloses a drilling and completion operation site environment risk early warning method, which belongs to the technical field of drilling engineering monitoring, and is characterized by comprising the following steps: S1, collecting drilling site production data and drilling site environment protection data; s2, collecting and analyzing image data through an artificial intelligence video analysis module; s3, performing data integration on the production data, the environmental protection data and the image data through a neural network model; and S4, the risk level is judged, and early warning is carried out. The production data, the environmental protection data and the image data are organically combined, data integration is carried out, and the early warning efficiency and the early warning accuracy can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of drilling engineering monitoring, and particularly relates to a method for early warning of on-site environmental risks in drilling and completion operations. Background Art

[0002] The environmental risks at the drilling and completion operation site mainly rely on manual identification and reporting. The accuracy of risk and hidden danger identification is related to the work experience and professional knowledge level of personnel, often resulting in untimely risk identification and emergency response.

[0003] Chinese patent document with publication number CN110121053A and publication date August 13, 2019 discloses a video monitoring method for risk classification and early warning at a drilling site, which is characterized in that it includes a collection module, a processing module, a determination module, a judgment module and an early warning module, wherein: The collection module is used to collect visual images of the drilling site; The processing module is used to analyze the visual images to obtain the characteristic information of the visual images and the semantic text information matching the characteristic information of the visual images; The determination module is used to determine the operation scenario according to the characteristic information of the video images and the semantic text information; The operation scenarios include the static state, constant speed, variable speed movement, hoisting or lowering of the drill pipe, the state of substances, and the operation scenarios corresponding to the semantic text information of safety belts, safety helmets, derricks, drill pipes, mud, second floors, and tongs; The judgment module is used to match the characteristic information of the video images and the semantic text information with the pre-loaded event templates to generate a mapping relationship database; The mapping relationship database includes the liquid level height of the mud pit, the on-site liquid level area and the flame area; The early warning module is used to give early warnings according to the risk levels; Specifically, it includes: Step a: Real-time collect visual images of the drilling site; Step b: Analyze the visual images to obtain the characteristic information of the visual images, and search for image semantics matching the characteristic information of the visual images in the scene model library to generate semantic text information; Step c: Determine the drilling site operation scenario according to the characteristic information of the visual images and the semantic text information; Step d: Match the characteristic information of the visual images and the semantic text information with the pre-loaded event templates to generate a mapping relationship database; Step e: Compare the mapping relationship database with historical data. If the matching degree is greater than the threshold, evaluate the risk of potential accidents and determine the event type according to the occurrence frequency and severity of historical similar accidents; The potential accident risk value is determined using a risk matrix; the risk levels of the potential accident risk value are divided into four levels: major risk, relatively large risk, general risk, and low risk; The risk index value of the major risk is 43 - 200, the risk index value of the relatively large risk is 20 - 37, the risk index value of the general risk is 10 - 17, and the risk index value of the low risk is 1 - 8; Each specific number in the risk matrix represents the risk index value of the risk, with the minimum being 1 and the maximum being 200; Step f: Conduct risk early warning according to the risk level. When the risk index value is greater than the set threshold, an alarm is issued; Each number in each specific risk level in the risk matrix represents the risk index value of the risk, and each absolute risk value can be converted into a risk index value.

[0004] The video monitoring method for risk classification and early warning at the drilling site disclosed in this patent document can conduct 24 - hour real - time risk classification monitoring and early warning on the video monitoring of important targets and operation activities at the drilling site, and can conduct safety risk classification assessment and provide early warning according to different situations. However, the early warning efficiency and accuracy are relatively low. Summary of the Invention

[0005] In order to overcome the defects of the above - mentioned prior art, the present invention provides a method for early warning of environmental risks at the drilling and completion operation site. By organically combining production data, environmental protection data, and image data and integrating the data, the present invention can effectively improve the early warning efficiency and accuracy.

[0006] The present invention is realized through the following technical solutions: A method for early warning of environmental risks at the drilling and completion operation site, characterized by including the following steps: S1. Collect production data and environmental protection data at the drilling site; S2. Collect and analyze image data through an artificial intelligence video analysis module; S3. Integrate production data, environmental protection data, and image data through a neural network model; S4. Judge the risk level and conduct early warning.

[0007] The step S2 specifically includes: S21. Image detection, identifying the image scene through a visual sensor, and the image scene includes Face recognition scene, photo recognition scene, and object recognition scene; S22. Feature extraction, extracting image features, including particle size, shallow - layer feature representation, structural feature representation, and feature dimension; S23. Deep learning, through multi-level non-linear information processing and abstraction, performs supervised or unsupervised feature learning, representation, classification, and pattern recognition; S24. Training dataset, collecting the ground datasets captured by on-site monitoring cameras and the ground datasets captured by drones and performing manual annotation, and identifying the dataset pictures; S25. Detection process, in image recognition, first finding the object position in advance, and then scanning the object position.

[0008] In the step S23, a convolutional neural network is used for deep learning for image processing of face recognition and object detection.

[0009] In the step S24, the dataset pictures include people, equipment, facilities, and roads.

[0010] In the step S25, finding the object position in advance specifically refers to learning and judging the theoretical existence position of the object according to the pre-entered data.

[0011] In the step S25, scanning the object position specifically refers to performing a scanning inspection according to the theoretically existing position of the object judged in advance.

[0012] In the step S3, data integration specifically refers to combining the environmental protection data and image data of each production stage with the current stage.

[0013] In the step S4, judging the risk level specifically refers to presetting a neural network model for judging the risk level, and then performing risk classification according to the data in the steps S2 and S3.

[0014] In the step S1, the drilling site production data is collected through the production data acquisition module.

[0015] In the step S1, the drilling site environmental protection data is collected through the environmental protection data acquisition module.

[0016] The beneficial effects of the present invention are mainly manifested in the following aspects: 1. Compared with the prior art, the present invention can effectively improve the early warning efficiency and accuracy by organically combining production data, environmental protection data, and image data and performing data integration.

[0017] 2. Compared with the prior early warning mainly based on manual subjective judgment, the present invention can make a more objective and accurate judgment based on work data, which is beneficial to improving the early warning effect.

[0018] 3. Based on artificial intelligence deep learning, the present invention can further improve the accuracy of evaluation by continuously optimizing the evaluation neural network model.

[0019] 4. In the present invention, in image recognition, it is necessary to find out which objects exist in an image, which is a many-to-many mapping. Compared with the existing convolutional neural network that first needs to find out which regions have objects and then use the convolutional neural network to classify them, by pre-finding the possible positions where objects may exist and then scanning these positions, the efficiency and accuracy of image recognition can be effectively improved.

[0020] 5. In the present invention, by pre-finding the object positions and then scanning these positions, compared with the prior art of scanning the entire image with sliding windows of different sizes, the amount of calculation can be effectively reduced and the efficiency of image recognition can be improved.

[0021] 6. The production data and environmental protection data of the drilling site obtained in the present invention can reflect the real-time state of the site and have better real-time performance.

[0022] 7. In the present invention, according to the collected data, the on-site environmental risks are evaluated in a timely manner and the environmental risks are classified, which has good timeliness.

[0023] 8. In the present invention, according to the classification of environmental risks, warnings of different levels can be issued, which is beneficial to ensuring the warning effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further specifically described below in conjunction with the drawings in the specification and the specific embodiments: Figure 1 It is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Embodiment 1 Refer to Figure 1 , a method for warning of on-site environmental risks in drilling and completion operations, comprising the following steps: S1. Collect the production data of the drilling site and the environmental protection data of the drilling site; S2. Collect and analyze image data through an artificial intelligence video analysis module; S3. Integrate the production data, environmental protection data and image data through a neural network model; S4. Judge the risk level and issue a warning.

[0026] This embodiment is the most basic implementation mode. Compared with the prior art, by organically combining the production data, environmental protection data and image data and integrating the data, the warning efficiency and warning accuracy can be effectively improved.

[0027] Embodiment 2 Refer to Figure 1 , a method for warning of on-site environmental risks in drilling and completion operations, comprising the following steps: S1. Collect the production data and environmental protection data at the drilling site; S2. Collect and analyze image data through the artificial intelligence video analysis module; S3. Integrate the production data, environmental protection data, and image data through the neural network model; S4. Judge the risk level and issue a warning.

[0028] The specific steps of step S2 include: S21. Image detection: Identify the image scene through a vision sensor. The image scene includes face recognition scene, photo recognition scene, and object recognition scene; S22. Feature extraction: Extract image features, including granularity, shallow feature representation, structural feature representation, and feature dimension; S23. Deep learning: Perform supervised or unsupervised feature learning, representation, classification, and pattern recognition through multi-level non-linear information processing and abstraction; S24. Training dataset: Collect the ground datasets captured by on-site surveillance cameras and drones and perform manual annotation, and identify the dataset pictures; S25. Detection process: In image recognition, first find the object position, and then scan the object position.

[0029] This embodiment is a preferred embodiment. Compared with the existing warning mainly based on manual subjective judgment, it can make a more objective and accurate judgment based on the working data, which is beneficial to improving the warning effect.

[0030] Embodiment 3 See Figure 1 , a method for early warning of on-site environmental risks during drilling and completion operations, including the following steps: S1. Collect the production data and environmental protection data at the drilling site; S2. Collect and analyze image data through the artificial intelligence video analysis module; S3. Integrate the production data, environmental protection data, and image data through the neural network model; S4. Judge the risk level and issue a warning.

[0031] Further, the specific steps of step S2 include: S21. Image detection: Identify the image scene through a vision sensor. The image scene includes face recognition scene, photo recognition scene, and object recognition scene; S22. Feature extraction: Extract image features, including granularity, shallow feature representation, structural feature representation, and feature dimension; S23. Deep learning, through multi-level non-linear information processing and abstraction, performs supervised or unsupervised feature learning, representation, classification, and pattern recognition; S24. Training dataset, collecting ground datasets captured by on-site monitoring cameras and ground datasets captured by drones and performing manual annotation, and identifying the dataset pictures; S25. Detection process, in image recognition, first finding the object position and then scanning the object position.

[0032] In the step S23, a convolutional neural network is used for deep learning for image processing in face recognition and target detection.

[0033] In the step S24, the dataset pictures include people, equipment, facilities, and roads.

[0034] This embodiment is another preferred embodiment. Based on artificial intelligence deep learning, by continuously optimizing the evaluation neural network model, the accuracy of the evaluation can be further improved.

[0035] Embodiment 4 See Figure 1 , a method for early warning of on-site environmental risks in drilling and completion operations, including the following steps: S1. Collecting drilling site production data and drilling site environmental protection data; S2. Collecting and analyzing image data through an artificial intelligence video analysis module; S3. Integrating production data, environmental protection data, and image data through a neural network model; S4. Judging the risk level and giving an early warning.

[0036] The step S2 specifically includes: S21. Image detection, identifying the image scene through a visual sensor, and the image scene includes face recognition scene, photo recognition scene, and object recognition scene; S22. Feature extraction, extracting image features, including granularity, shallow feature representation, structural feature representation, and feature dimension; S23. Deep learning, through multi-level non-linear information processing and abstraction, performs supervised or unsupervised feature learning, representation, classification, and pattern recognition; S24. Training dataset, collecting ground datasets captured by on-site monitoring cameras and ground datasets captured by drones and performing manual annotation, and identifying the dataset pictures; S25. Detection process, in image recognition, first finding the object position and then scanning the object position.

[0037] Further, in step S23, deep learning is performed using a convolutional neural network for image processing for face recognition and object detection.

[0038] In step S24, the dataset pictures include people, equipment, facilities, and roads.

[0039] In step S25, finding out the object position in advance specifically refers to learning and judging the theoretical existence position of the object based on the pre-entered data.

[0040] In step S25, scanning the object position specifically refers to performing a scanning inspection based on the theoretically existing position of the object judged in advance.

[0041] This embodiment is another preferred embodiment. In image recognition, it is necessary to find out which objects are in an image, which is a many-to-many mapping. Compared with the existing convolutional neural network that needs to first find out which areas have objects and then use the convolutional neural network to classify them, by finding out the possible positions where the objects may exist in advance and then scanning that position, the efficiency and accuracy of image recognition can be effectively improved.

[0042] Embodiment 5 See Figure 1 , a method for early warning of on-site environmental risks in drilling and well completion operations, including the following steps: S1. Collect production data and environmental protection data at the drilling site; S2. Collect and analyze image data through an artificial intelligence video analysis module; S3. Integrate production data, environmental protection data, and image data through a neural network model; S4. Judge the risk level and give an early warning.

[0043] Step S2 specifically includes: S21. Image detection, identifying the image scene through a vision sensor, and the image scene includes face recognition scene, photo recognition scene, and object recognition scene; S22. Feature extraction, extracting image features, including granularity, shallow feature representation, structural feature representation, and feature dimension; S23. Deep learning, performing supervised or unsupervised feature learning, representation, classification, and pattern recognition through multi-level non-linear information processing and abstraction; S24. Training dataset, collecting the ground dataset captured by on-site monitoring cameras and the ground dataset captured by drones and performing manual annotation, and identifying the dataset pictures; S25. Detection process, in image recognition, finding out the object position in advance and then scanning the object position.

[0044] In step S23, deep learning is performed using a convolutional neural network for image processing of face recognition and object detection.

[0045] In step S24, the dataset pictures include people, equipment, facilities, and roads.

[0046] Furthermore, in step S25, finding out the specific position of the object in advance specifically refers to learning and judging the theoretical existence position of the object according to the pre-entered data.

[0047] In step S25, scanning the position of the object specifically refers to performing a scanning inspection according to the theoretically existing position of the object judged in advance.

[0048] In step S3, data integration specifically refers to combining the environmental protection data and image data of each production stage with the current stage.

[0049] In step S4, judging the risk level specifically refers to presetting a neural network model for judging the risk level, and then performing risk classification according to the data in steps S2 and S3.

[0050] In step S1, the production data of the drilling site is collected through the production data acquisition module.

[0051] In step S1, the environmental protection data of the drilling site is collected through the environmental protection data acquisition module.

[0052] This embodiment is the best implementation method. By finding out the position of the object in advance and then scanning the position of the object, compared with the prior art of scanning the entire image with sliding windows of different sizes, it can effectively reduce the calculation amount and improve the image recognition efficiency.

[0053] The obtained production data of the drilling site and the environmental protection data of the drilling site can reflect the real-time state of the site and have better real-time performance.

[0054] Evaluating the on-site environmental risk in a timely manner according to the collected data and classifying the environmental risk have good timeliness.

[0055] According to the classification of the environmental risk level, different levels of early warnings can be issued, which is beneficial to ensuring the early warning effect.

[0056] The principle of the present invention is as follows: Collect the production data and environmental protection data of the drilling site. Use the artificial intelligence video analysis module to collect and analyze image data. In image recognition, it is necessary to find out which objects are in an image, which is a many-to-many mapping. Compared with the existing convolutional neural network that needs to first find out which areas have objects and then use the convolutional neural network to classify them, by pre-finding the possible positions where objects may exist, then scanning that position, and finally combining and analyzing the environmental protection data and image data at each stage of production through the neural network model, the image recognition efficiency and accuracy can be effectively improved.

Claims

1. A method for early warning of on-site environmental risks in drilling and well completion operations, characterized in that, It includes the following steps: S1. Collect the production data and environmental protection data at the drilling site; S2. Collect and analyze the image data through the artificial intelligence video analysis module; S3. Integrate the production data, environmental protection data and image data through the neural network model; S4. Judge the risk level and give an early warning.

2. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 1, characterized in that: The step S2 specifically includes: S21. Image detection, identify the image scene through the visual sensor, and the image scene includes face recognition scene, photo recognition scene and object recognition scene; S22. Feature extraction, extract image features, including particle size, shallow feature representation, structural feature representation and feature dimension; S23. Deep learning, through multi-level non-linear information processing and abstraction, perform supervised or unsupervised feature learning, representation, classification and pattern recognition; S24. Training data set, collect the ground data set captured by the on-site monitoring camera and the ground data set captured by the drone and perform manual annotation, and identify the pictures in the data set; S25. Detection process, in image recognition, first find out the object position in advance, and then scan the object position.

3. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 2, characterized in that: In the step S23, a convolutional neural network is used for deep learning for image processing of face recognition and target detection.

4. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 2, characterized in that: In the step S24, the data set pictures include people, equipment, facilities and roads.

5. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 2, characterized in that: In the step S25, finding out the object position in advance specifically means learning and judging the theoretical existence position of the object according to the pre-entered information.

6. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 2, characterized in that: In the step S25, scanning the object position specifically means scanning and checking according to the pre-judged theoretical existence position of the object.

7. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 1, characterized in that: In the step S3, data integration specifically means combining the environmental protection data and image data of each production stage with the current stage.

8. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 1, characterized in that: In the step S4, judging the risk level specifically means presetting a neural network model for judging the risk level, and then grading the risk according to the data in steps S2 and S3.

9. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 1, characterized in that: In the step S1, the production data at the drilling site is collected through the production data collection module.

10. The method for early warning of on-site environmental risks in drilling and well completion operations according to claim 1, characterized in that: In the step S1, the environmental protection data at the drilling site is collected through the environmental protection data collection module.

Images